Background  The proportion of severe patients in the early stage of COVID-19 is high, and its clinical characteristics and prognostic risk factors need to be further clarified.

  Objective  To retrospectively analyze the clinical features and prognostic risk factors of severe COVID-19 patients.

  Methods  A total of 319 patients with severe COVID-19 in 2020 were selected as research objects. Baseline clinical characteristics, laboratory test indicators, CT imaging characteristics and clinical scores of the patients were analyzed. The differences between survived and deceased patients and mortality related risk factors were analyzed.

  Results  Among 319 cases of severe disease, 262 cases were severe and 57 cases were critical, 11 severe cases and 35 critical cases died during hospitalization. The median age of the dead patients, the proportion of symptoms such as fever, the proportion of underlying diseases, and the proportion of comorbidities such as ARDS and abnormal coagulation were all higher than those of the survived patients (P<0.05). The main CT manifestations of survived patients were patch shadow and ground glass shadow, and the proportions of realistic shadow and pleural effusion on lung CT of deceased patients were higher than those of survived patients (P<0.05). Admission examination showed that the median CRP and D-dimer were increased and the median lymphocyte count was decreased in 319 severe patients, but the degree of increase or decrease in deceased patients was more significant than that in survived patients (P<0.05). Multivariate logistic analysis suggested that ARDS during hospitalization (OR=7.904, 95% CI: 1.899-32.907) was associated with adverse outcomes. The higher the age (OR=1.054, 95% CI: 1.005-1.105), qSOFA score (OR=3.787, 95% CI: 1.616-8.874), CRP (OR=1.013, 95% CI: 1.004-1.022) and D-dimer (OR=1.104, 95% CI: 1.006-1.212), the greater the risk of death (P<0.05).

  Conclusion  Complicated with underlying diseases, comorbidities, abnormal infection index and abnormal coagulation index are associated with death, which need to be identified early. Age, CRP, D-dimer and qSOFA scores can be used as predictors to assess the risk of death in hospital.